python遗传算法的优化过程是什么？

# 初始种群

pop = [Robot() for x in range(pop_size)]

results = []

# 执行进化

for i in tqdm(range(num_gen)):

    scores = np.zeros(pop_size)

    # 遍历所有机器人

    for idx, rob in enumerate(pop):

        # 运行垃圾收集模拟并计算拟合度

        score = rob.simulate(iter_per_sim, moves_per_iter)

        scores[idx] = score

    results.append([scores.mean(),scores.max()]) # 保存每一代的平均值和值

    best_robot = pop[scores.argmax()] # 保存最好的机器人

    # 限制那些能够交配的机器人的数量

    inds = np.argpartition(scores, -num_breeders)[-num_breeders:] # 基于拟合度得到顶级机器人的索引

    subpop = []

    for idx in inds:

        subpop.append(pop[idx])

    scores = scores[inds]

    # 平方并标准化

    norm_scores = (scores - scores.min()) ** 2

    norm_scores = norm_scores / norm_scores.sum()

    # 创造下一代机器人

    new_pop = []

    for child in range(pop_size):

        # 选择拟合度优秀的父母

        p1, p2 = np.random.choice(subpop, p=norm_scores, size=2, replace=False)

        new_pop.append(Robot(p1.dna, p2.dna))

pop = new_pop